This content is part of the Essential Guide: The rise of hyper-converged architecture for storage

Deconstructing hyperconverged infrastructure risks and rewards

Hyperconverged infrastructure helps break down data center silos, but it comes with personnel and vendor lock-in issues, and it might not be the right fit for every environment.

Hyperconverged infrastructure has raised eyebrows and piqued interest among IT professionals, but there's confusion around what it is, what risks it carries and which shops should consider it.

Search the Internet a bit and you'll find there's some uncertainty around hyperconvergence. In fact, you'll find that there doesn't even seem to be a consensus on the spelling.

In concept, hyperconverged infrastructure is a great idea. It eliminates silos in the data center and uses a single platform to provision and manage disparate data center services such as compute, storage and networking. Sounds a lot like the software-defined data center (SDDC), right? I've talked to IT professionals about this and their answers about whether hyperconverged infrastructure and SDDC are the same vary, but there are strong arguments on both sides. My answer is that they're not the same thing.

In this market, there are three sort-of similar categories. There's converged infrastructure, which is essentially a hardware-only way to provide compute, storage and switching. It's a dense platform with fewer bottlenecks than the traditional model. Then there's the SDDC, which in its purest form is hardware agnostic. SDDC is about moving the logical configuration from the hardware layer to the software layer so that it can be defined -- or redefined -- on the fly, without modifications to the underlying platform. The combination of converged infrastructure and SDDC is what we now refer to as hyperconverged infrastructure.

When shops say, 'We're a Cisco shop' they're saying, 'We're not interested in innovation. We'd rather maintain homogeneity in the network.'

It's a hardware platform that delivers compute, storage and networking in one, plus a management system and the ability for the software layer to define the operational aspects of that infrastructure.  Hyperconverged infrastructure is converged infrastructure where the operational aspects and configuration are managed by an intelligent (and often automated) SDDC.

By my definition, something like the Dell PowerEdge FX isn't hyperconverged until it's running a hypervisor with an orchestration element that handles the provisioning of storage, compute and networking resources. Likewise, an environment that's deployed as an SDDC using disparate hardware platforms isn't hyperconverged.

Well-known vendors such as Dell, Hewlett-Packard, Cisco and VMware have cooperative packages that deliver hyperconvergence. But the hyperconverged infrastructure market is growing every day, and it's not just dominated by big players. Vendors such as SimpliVity and Nutanix use hyperconvergence to blur the lines between on-premises infrastructure and the cloud.

You don't have to think too hard to see there could be big benefits for both Capex and Opex when it comes to hyperconvergence. But two big questions remain: What are the risks and where does it fit?

What are the risks?

In large-scale enterprise IT shops, siloes in the data center can create problems. Often, there are separate IT teams to handle storage, networking, virtualization, x86 and Unix, and some companies may still have a mainframe team. Using hyperconverged infrastructure can bring all those people under one umbrella, but you'll still have to contend with figuring out who's in charge. For shops with lots of separate groups each with their own management hierarchy, that can present a tough personnel issue.

Then there's the vendor silo problem. To truly get hyperconverged infrastructure, you are often pushed into selecting a single vendor platform, but sticking to one vendor for everything doesn't always allow you to take advantage of innovative new tools and products.

I'm an advocate for best-of-breed products, and I think vendor loyalty is counter-productive in the data center. When shops say, "We're a Cisco shop" they're saying, "We're not interested in innovation. We'd rather maintain homogeneity in the network." Personally, I don't want to be locked into a vendor at the expense of innovation.

Even for people who think I'm completely wrong about being open to multi-vendor options, it's important to consider whether deploying offerings such as SimpliVity's or VMware's EVO:RACK and RAIL lock you into specific hardware options.

Where does it fit?

I really like the idea of hyperconverged infrastructure. I think it offers a clear path to a hybrid infrastructure that allows workloads to scale dynamically and flex to and from the cloud as needed. To me, there's no better path for building an agile IT infrastructure.

But I also think hyperconvergence could be limiting in the core infrastructure of a company. And the right or best use cases for hyperconverged infrastructure have yet to be determined. Many companies with specialized workloads benefit from using specific platforms for those workloads. Those may or may not fit into the hyperconverged model (but would definitely fit in the SDDC).

Hyperconverged infrastructure is a good fit for companies with branch locations or colocated facilities, as well as small organizations with big infrastructure requirements. And why wouldn't you want 4U of gear that you can screw into a rack at the branch, power it on and be good to go?

Hyperconvergence will continue to evolve, and as it does, it will likely take on a multi-vendor posture as a function of consumer demand. Until that happens, I believe it's a powerful tool that must be applied only after serious consideration.

Next Steps

How well do you know EVO:RAIL and EVO:RACK?

Explore SimpliVity’s OmniCube

Nutanix and Dell release XC platform

Top 10 reasons to go hyper-converged

Dig Deeper on Virtual desktop infrastructure and architecture